Ready Player 3: The Rise of AI-Driven Development and the New Era of Product Engineering

Welcome to the game...

Product managers and software developers face an event horizon. Each group has powerful new AI tools that equip them to excel at tasks previously in the other group's domain. Product managers can use AI to become proficient–even skilled–coders. Software developers likewise have AI tools to help them become gifted product managers and marketers. 

Because of these newfound capabilities, advocating for the continued separation of traditional product management from software development makes less and less sense. We should call on product managers to get closer to the technology and software developers to cozy up to customers and users.

Will this event horizon occur in 2025? It’s hard to say, but it’s clear that segregating product management from software development roles and responsibilities diminishes a product organization’s ability to deliver impactful products. Simply put, our traditional views of product management, software development, and the craft of building software products seem antiquated in the AI era.

AI is not just reshaping the tech industry; it is revolutionizing how we consume software, changing the types of software teams build, and transforming traditional software lifecycles. Non-technical users now have the power to create software using tools like Cursor, ChatGPT, and Claude, enabling them to build simple applications in minutes with a single prompt. This capability allows them to create software on-demand, use it temporarily, and discard it once it's no longer needed. As a result, traditional software development teams can focus on empowering anyone to build and publish their software, marshaling their efforts around enabling platforms like backstage.io or finding ways to host models and apps users build. 

In June 2024, the ebook Situational Awareness a visionary goal for the AI industry: the creation of the first virtual senior AI engineer—a system capable of recursively improving itself and transforming the landscape of digital innovation. This bold objective isn’t a distant dream; it’s already taking shape in the transformative tools we see today. AI-driven platforms like GitHub Copilot and Cursor are no longer mere aids to developers; they’re beginning to take on integral responsibilities, shaping the way products are conceived, built, and iterated.

As AI increasingly takes on strategic, creative, and developmental roles within product organizations, we stand at the dawn of an AI-driven industry. This shift sets the stage for an era where advanced AI systems might reduce or eliminate the need for human oversight in key aspects of product creation. By empowering engineers with tools that autonomously iterate and enhance, we move closer to a future of self-sustaining, self-innovating AI—an industry where recursive advancement becomes a reality.

The cumulative effect of these changes is a new era of software development: Product Engineering—a crucial step as we prepare for AI systems to transform not only development processes but the nature of innovation itself.

What Is Product Engineering?

Product Engineering completely reimagines the software product development lifecycle. Traditionally, product managers would discover a user need, turn that into requirements, build a backlog of stories, and hand that off to engineers to develop and deliver the software. For a team with a large backlog, there’s no telling where the feature lands in the priority queue. This model isn’t great for end users because they have long waits for software that might accomplish an important but simple task. 

Now, users can skip the line. Anyone subscribed to a frontier LLM model can create and run code that accomplishes that simple task. Say I have a spreadsheet or database with some data I want to analyze, display the output of that analysis, and interact with the output in a simple dashboard. I might still be waiting on the software if I submitted that use case two years ago. Now, I can build a data analysis dashboard in minutes. Think of this as a gateway drug. Building more complex LLM agents is just a step or two beyond implementing these simple use cases.

With end users empowered to develop their software to satisfy simple needs, development teams can focus on more complex development tasks, building the tools that enable others to create their software rather than spending time on applications that have short lifecycles. Using tools like GitHub Copilot helps engineers code faster, with higher quality, and iterate in real-time based on feedback. GitHub’s CEO predicts that AI coding assistants like Copilot will soon develop 80% of code. If his prediction is accurate, and machines create the vast majority of products, why would you restrict who can ask the machine to do a task? You wouldn’t. In this model, product development's what, why, how, and when collapses into a single, seamless process.

The Obsolescence of Traditional Product Management

Traditional product management frameworks, like the model proposed in Marty Cagan’s Inspired, emphasized the separation between product strategy and execution. At the time, the distinction was important because it helped us manage and deploy team members with specialized skillsetsToday’s AI-driven development tools dissolve the distinction. The modern Product Engineer has become both the strategic architect and the code creator, delivering faster, more efficient results using  AI tools. This convergence reduces the back-and-forth between Product Managers and developers and minimizes the need for traditional documentation like PRDs or backlogs of stories. Teams still need business cases, go-to-market plans, and operating models. However, as coding spreads to non-technical team members, AI tools also help software developers operate outside their traditional skill sets.

While Product Management and Software Engineering are not disappearing, their roles blend. The long-standing goal of bringing product managers closer to the code and software engineers closer to customer insights is now achievable through AI. Product managers can prototype and write code themselves, turning traditional tools like Jira into platforms focused on optimizing delivery flow. Meanwhile, engineers can leverage AI to simulate customer interactions, embedding real-time feedback loops into their development processes—transforming how both roles collaborate in creating products.

We’re already seeing new roles emerge, like "Design Engineers," who bridge the gap between design and feature development, creating beautiful and technically sound​products. Similarly, Product Engineering will require engineers who know how to code and understand UX, customer empathy, and business strategy. 

The End of Support, QA, and DevOps as We Know Them

Imagine a world where product support is obsolete. Support exists to bridge the gap between identifying defects and implementing fixes, but AI can close that gap faster than ever. Product Engineers will use AI tools to find and resolve software issues proactively, quickly deploy fixes, and delegate responses to support tickets to AI agents, effectively eliminating the need for a separate support team.

QA teams are experiencing a steady decline, accelerated by AI-driven automation that can analyze requirements, define test cases, and execute automated testing with minimal human oversight. Beyond QA, AI agents are now managing critical infrastructure platforms like compute, data storage, and networking, where they reduce human error in configuration, setup, and security vulnerabilities—common causes of cybersecurity breaches. Issues such as leaving S3 buckets open to the world or falling prey to phishing attacks are minimized with AI taking charge of these platforms. As AI becomes the primary builder and maintainer, the need for human administrators will continue to dwindle, resulting in safer, more resilient systems.

One of the most compelling implications of AI-managed infrastructure is its potential to reduce cybersecurity risks significantly. Human errors, such as misconfiguring systems or falling victim to phishing, are often the weak link in cybersecurity. With AI agents managing these platforms and even acting as the ultimate users, the surface area for human-induced errors diminishes. This shift could mark a fundamental change in how organizations approach security—automating trust by allowing AI to oversee infrastructure and software, ensuring tighter control and faster response to potential threats.

The Offshore/Nearshore Development Model Is Kaput

One of this shift's most profound side effects will be the end of the offshore/nearshore development model as we know it. Traditionally, companies have relied on outsourced development teams for cost savings and scalability. However, organizations will increasingly replace offshore developers with AI developers as AI becomes more advanced. These AI-driven tools will perform much of the coding work currently outsourced, operating under the supervision of senior Product Engineers.

This shift won’t just impact the companies that have historically outsourced development; outsourced development partners will also need to evolve. Their role is likely to pivot toward specializing in legacy code migration projects, where they will leverage AI to help modernize outdated systems. In this new model, they won’t just be coding; they will be using AI to inspect, understand, and refactor legacy systems into modern frameworks. This natural evolution will allow outsourcing firms to remain relevant in a world where routine coding work is increasingly automated.

Flattening Organizational Structures

The definition of a "product organization" is evolving beyond traditional silos to include teams like Sales, Marketing, Customer Success, Legal, and Finance. With AI tools, these non-technical teams are direct participants in the guts of the product development lifecycle. A sales engineer or product marketer can learn of a requirement, prototype it, and solicit user feedback before engaging a product manager or software developer.

Most organizations manage the specialization of skills by creating siloed and deeply hierarchical structures. Modeling people's relationships like this is a hangover from the Industrial Revolution. There’s no reason to do it like this in the AI age. AI enhances and redistributes skills making these silos and hierarchies obsolete. With all due respect to our marketing friends, AI has already mastered the basics of their jobs.

Deploying tactics like domain-driven design aligns the entire organization by providing a unified language for building, positioning, and selling products, a task easily delegated to AI agents. Organizations can reallocate budgets that once funded multi-tiered product teams toward innovation and scaling. With Product Engineers leading the charge, organizations will become leaner, more flexible, and faster at delivering value to customers.

One unintended consequence is that many companies are becoming independent from traditional tech providers. Companies like Oracle and Microsoft lose their grip on the market when anyone at any company can develop software that potentially replaces expensive enterprise software at a fraction of the cost. Microsoft is already redeploying massive amounts of capital on AI technology and its underlying infrastructure, which allows customers to displace their traditional cash cows. Tech companies unable to adjust their strategies and adapt to these new market realities will fail.

The Emergence of New Roles and Curriculums

To prepare for this new reality, we must rethink how we educate and train Product Engineers. Most traditional university programs treat software development and product management as separate disciplines. Not surprisingly, companies hire and compensate employees based on their educational specialization. 

Some universities have adapted by developing master's programs that combine a traditional MBA with a master's in computer science. Northeastern University, Carnegie Mellon University, and the University of Chicago–along with a host of other innovative schools–have developed such programs. Hiring companies aren’t yet taking advantage of the combined skillset. Employers still track 64% of students who graduated from the Chicago Booth School of Business program into software development roles. 

Companies need to make space for these cross-disciplinary executives to land. For how much longer does it make sense to have technical organizations headed by a Chief Technology Officer as separate and distinct from product or marketing organizations headed by a Chief Product Officer or Chief Marketing Officer? What about the commercial and business operations organizations? The fact that AI takes on tasks that require specialization–even deeply specialized skills like software engineering, technology operations, legal, and finance–drives speculation that the first single-person company valued at $1B isn’t far off.

Repaying Technical Debt in Real-Time

AI is also revolutionizing how we manage infrastructure platforms like compute, data storage, and networking, minimizing human error and maximizing operational efficiency. AI agents execute tasks like server setup, network configurations, and platform maintenance more accurately, which can automate error detection, resolve misconfigurations, and optimize performance continuously. Delegating infrastructure management to AIs significantly reduces the human-induced errors that often cause security issues, allowing companies to focus on innovation and growth rather than battling security lapses or operational mistakes.

With AI tools like Cursor, Product Engineers can now address and repay technical debt as part of their daily workflow. Product Engineers create backlogs undiluted by technical debt, keeping features as the team’s highest priority and helping the team be productive and efficient. Additionally, "technical debt foreclosure" can be introduced to signify when a product has become impossible or cost-prohibitive to maintain. In such cases, it’s essential to recognize when a platform needs to be sunset and replaced with a new implementation on an updated tech stack.

AI can play a crucial role in this process, understanding the legacy code and making it feasible to transition smoothly to a new platform. With insights driven by AI, technical debt foreclosure becomes manageable, allowing companies to process this debt to ensure sustainable, long-term growth while keeping platforms agile and up-to-date.

Legacy Code Migration: AI’s Role in Refactoring

AI is poised to revolutionize legacy code migration. While tools like Amazon CodeGuru can already analyze and interpret legacy code, the future lies in AI, not merely converting outdated languages like COBOL into modern languages such as Node.js. Instead, Product Engineers can ask AI to comprehensively analyze and understand the logic and intentions behind legacy code, document the product’s foundational specifications, and refactor it into modern frameworks only after review and approval.

This AI-driven approach to legacy code refactoring offers more than a direct translation—it introduces the opportunity to enhance, streamline, and elevate existing products to preserve functionality while eliminating historical inefficiencies. Through AI-supported analysis, legacy systems can undergo transformations that make code leaner and more maintainable by addressing unnecessary complexity and reducing feature bloat. By removing or optimizing outdated, redundant code elements, products become less cumbersome, delivering faster performance and lowering maintenance costs.

This process is a rare chance to improve user experience and create fresh value. Streamlined code means more responsive and user-friendly applications, which can directly impact user satisfaction and retention. Additionally, the migration process allows teams to re-evaluate and refine features based on current user needs, consolidating functionalities that might have accumulated over the years but no longer serve their original purpose. In this way, AI doesn’t just carry forward the value of legacy code—it transforms it, enabling a simpler, optimized product with an elevated user experience and a more future-proof foundation.

The Fall (or Evolution) of Product Management Institutes

So, where does this leave the giants of Product Management education, like the Product Management Institute or Pragmatic Marketing? Well, either they adapt, or they face extinction. Perhaps they’ll evolve into centers of excellence for Product Engineering—offering new courses that teach aspiring Product Engineers how to manage a product and build one from the ground up. After all, even dinosaurs had to adapt to survive. These institutions must teach product strategy, how to code, empathize with users, and deliver working software in a continuous loop.

Can you imagine a "Certified Product Engineer" badge from the PMI? Well, if they want to stay relevant, they better start printing those certifications now.

The Rise of Ephemeral Software

We build software the way we do because not everyone has the skills to do it. The economics of relying on specialists to build software means keeping that software around long enough to return the investment required to create it. We’ve devised ways to deliver long-lived software and then struggled to make delivery faster, cheaper, and more adaptive to user needs and new requirements. Improvements have been mostly incremental, like making the software development processes “agile” and renting our computers from someone else’s data centers.

AI gives everyone the skills to develop software. If I have a specific problem, I can ask AI to build a tool to solve it, and the AI will implement the solution more cheaply and faster than humans have ever produced software. I may need to use that tool every day, or I may need to modify that tool slightly for different tasks. More often, I only need that tool to perform a task once.

When we measure feature usage of traditional, long-lived software products, we know this is typical behavior. Most software suffers from the Pareto Principle, and 80% of features are rarely or never used. Software development teams have been fighting this for decades, but we keep building software because we get compensated to build software.

Now that end users can build their own software, it’s ephemeral.

Packaged software delivered as a service at high per-user / per-month costs–for example, marketing automation or data analytics–is about to become relatively low value. The future isn’t Salesforce or Hubspot; it’s Hugging Face. The most successful companies of the next decade will be those building the frontier models, the data centers to train and run them, or the frameworks, like Cursor, that help people use the models to build ephemeral software.

AI and ephemeral software will replace so-called modern software much faster than modern software replaced mainframes. Companies deeply embed products like ServiceNow into their operations. But AI-developed, ephemeral software executes a strangler pattern organically, quickly choking off all the rarely used features and leaving vendors dreading the rise of license utilization tools. 

As software becomes more on-demand and disposable, traditional support and maintenance structures give way to AI-driven, self-sustaining systems. These systems autonomously detect and resolve issues, perform updates, and ensure smooth operations with minimal human oversight. This transition frees development teams to focus on innovation and platform growth while AI manages the lifecycle of these temporary, purpose-driven software solutions. In this new ecosystem, software is adaptive and customizable, existing only as long as necessary, transforming how we build, use, and support applications.

The Cambrian Explosion of Innovation

The best part? This evolution opens new doors. Efficiency and productivity will reach levels we've never seen before. The barriers to entry for building successful products will shrink dramatically, enabling solopreneurs and two-person teams to develop products that can compete with massive organizations. This shift will drive an explosion of innovation, ingenuity, and human expression—a Cambrian explosion of creativity where anyone with a vision can execute it without bloated teams or endless approvals.

As non-tech companies become more independent from traditional tech giants like Oracle and Microsoft, the software development landscape is shifting. Companies can now create, test, and deliver code on their own without being constrained by the centralized structures of large tech providers. This "democratization of software development" is not just a buzzword; it’s reshaping the market. SaaS, data storage, and analytics sectors are seeing a collapse in their traditional models while new areas focused on citizen-developed software are emerging. This shift enables faster iterations, better alignment with specific business needs, and the ability for smaller teams to compete with much larger organizations.

What we’re witnessing isn’t just a change in the job descriptions of engineers and PMs—it’s the dawn of a new era for product creation. The future belongs to the creators, not just the coders. In this AI-driven era, anyone with a vision can bring software to life, use it when needed, and discard it once its purpose is fulfilled. This shift democratizes innovation and accelerates it as software becomes more fluid, on-demand, and temporary than ever before. The era of Product Engineering has begun, and the potential for transformation is limitless.

The future belongs to Product Engineers who can strategize, design, code, deliver, market, sell, and maintain software, build and nurture lasting customer relationships, and create genuinely impactful products.